The US AI landscape is defined by a massive valuation gap that infrastructure builders need to understand. Nine American companies, including OpenAI, Anthropic, and xAI, hold a combined valuation of $2.18 trillion, while their total 2025 revenue sits at just $22.2 billion. This disparity, highlighted in a new market structure brief, reveals that investors are betting on revenue slopes rather than current financial health. For developers and infrastructure teams, this means the tools you rely on are being propped up by capital flows that often loop back to the hardware and cloud providers supplying the compute.

The Revenue vs. Valuation Disconnect

Anthropic’s recent SEC filing illustrates the extreme multiples at play. The company raised $65 billion at a $965 billion valuation in May 2026, despite reporting 2025 revenue of only $4.59 billion. This equates to roughly $210 in valuation for every dollar earned. However, the company’s run-rate revenue has since surged to $47 billion, driven by first-half 2026 sales of $16.29 billion. This rapid growth justifies the high multiple only if the steep trajectory continues, masking the fact that compute costs alone consumed $7.33 billion of Anthropic’s $12.65 billion in operating expenses last year.

Circular Funding and Supplier Dependencies

A closer look at the investor lists reveals a circular economy where suppliers finance their own customers. NVIDIA invested $30 billion in OpenAI and co-led rounds for Thinking Machines, while also selling hardware to all major labs. Amazon committed $50 billion to OpenAI and holds convertible notes in Anthropic, while simultaneously selling Anthropic $110 billion in compute capacity over the next decade. Google has similarly invested in Anthropic while providing cloud infrastructure. This setup mirrors the late-1990s telecom bubble, where equipment vendors lent money to carriers to buy their products, creating a fragile dependency if end-user revenue fails to materialize.

The Ex-OpenAI Talent Pipeline

Three of the nine major labs—Anthropic, Safe Superintelligence, and Thinking Machines—were founded by former OpenAI employees. Dario and Daniela Amodei left to start Anthropic in 2021, while Ilya Sutskever founded SSI in 2024 and Mira Murati launched Thinking Machines in 2025. This concentration of talent means that a significant portion of the field’s practical knowledge is held by a small group of people who trained together. For infrastructure teams, this suggests that technical roadmaps across these competitors may converge, as founders rely on shared insights from their time at OpenAI rather than entirely novel approaches.

Key Takeaways

  • Nine US AI labs hold a $2.18 trillion valuation against only $22.2 billion in 2025 revenue.
  • Anthropic’s valuation multiple dropped from 210x to 21x as run-rate revenue surged to $47 billion.
  • Major investors like NVIDIA, Amazon, and Google are also primary compute suppliers, creating circular funding dependencies.
  • Three leading labs were founded by ex-OpenAI staff, indicating a concentrated talent pool driving industry standards.
  • xAI’s $250 billion valuation is an internal SpaceX merger price, not an external market mark.

The Bottom Line

The AI boom is currently an infrastructure subsidy disguised as a software revolution, where cloud and chip giants are effectively funding their own customers to keep the demand curve steep. Builders should be wary of a market correction if these circular capital flows stop, as the underlying unit economics remain unproven.